Bipartite and tripartite Cucumber mosaic virus-based vectors for producing the Acidothermus cellulolyticus endo-1,4-β-glucanase and other proteins in non-transgenic plants

Background Using plant viruses to produce desirable proteins in plants allows for using non-transgenic plant hosts and if necessary, the ability to make rapid changes in the virus construct for increased or modified protein product yields. The objective of this work was the development of advanced CMV-based protein production systems to produce Acidothermus cellulolyticus endo-1, 4-β-glucanase (E1) in non-transgenic plants. Results We used two new Cucumber mosaic virus (CMV)-based vector systems for producing the green fluorescent protein (GFP) and more importantly, the Acidothermus cellulolyticus endo-1, 4-β-glucanase (E1) in non-transgenic Nicotiana benthamiana plants. These are the inducible CMVin (CMV-based inducible) and the autonomously replicating CMVar (CMV-based advanced replicating) systems. We modified a binary plasmid containing the complete CMV RNA 3 cDNA to facilitate insertion of desired sequences, and to give modifications of the subgenomic mRNA 4 leader sequence yielding several variants. Quantitative RT-PCR and immunoblot analysis showed good levels of CMV RNA and coat protein accumulation for some variants of both CMVin and CMVar. When genes for E1 or GFP were inserted in place of the CMV coat protein, both were produced in plants as shown by fluorescence (GFP) and immunoblot analysis. Enzymatic activity assays showed that active E1 was produced in plants with yields up to ~ 11 μg/g fresh weight (FW) for specific variant constructs. We also compared in vitro CMV genomic RNA reassortants, and CMV RNA 3 mutants which lacked the C’ terminal 33 amino acids of the 3A movement protein in attempts to further increase E1 yield. Taken together specific variant constructs yielded up to ~21 μg/g FW of E1 in non-transgenic plants. Conclusions Intact, active E1 was rapidly produced in non-transgenic plants by using agroinfiltration with the CMV-based systems. This reduces the time and cost compared to that required to generate transgenic plants and still gives the comparable yields of active E1. Our modifications described here, including manipulating cloning sites for foreign gene introduction, enhance the ease of use. Also, N. benthamiana, which is particularly suitable for agroinfiltration, is a very good plant for transient protein production.


Background
Using plant viruses as vehicles for foreign protein production in plants offers many advantages over more traditional prokaryotic-based, and even over transgenic plant-based protein production systems. For example, plants are relatively easy and inexpensive to grow, plants are able to perform post-translational protein modifications (e.g. glycosylation) not possible with prokaryotes, and plant cells can secrete appropriately engineered proteins allowing for simplified product purification [1,2]. Transgenic plants engineered to produce desirable proteins offer some of these advantages, but engineered plants require substantial time, effort and cost to develop and do not offer flexibility for rapid change if modifications to the protein product are desired. By contrast, using plant viruses to produce desired proteins in plants allows for using non-transgenic plant hosts and if necessary, the ability to make rapid changes in the virus construct for increased or modified protein product yields.
Cucumber mosaic virus (CMV) is the one of the viruses that has been used for protein production in plants [3,4]. CMV has an extremely wide plant host range [5] which opens the door for using plants other than only Nicotiana spp. for producing proteins, and thus an optimized CMV-based protein production system would be very desirable. But CMV also has some potential drawbacks for foreign protein production. CMV has a tripartite single-stranded RNA genome and each genomic RNA is packaged separately within icosahedral capsids [5]. CMV genomic RNAs 1 and 2 encode the 1a and 2a proteins, respectively, which are involved in viral RNA replication [5,6]. RNA 2 also encodes a small protein called 2b, which affects virulence and is known to suppress the initiation of the plant defense, RNA silencing, and to play a role in promoting cellto-cell movement [7]. RNA 3 also is bicistronic, encoding the cell-to-cell movement protein (MP) and the virion capsid protein (CP). All three CMV genomic RNAs are essential for the systemic plant infection and all five CMV-encoded proteins directly or indirectly affect the movement of CMV within the plant host [5]. Still, CMV genome segments 2 and 3 have been modified in some cases for insertion of specific sequences which can give foreign protein production in plants [4,8,9].
In our previous work we engineered a binary plasmid to contain modified complementary DNAs (cDNAs) representing the complete CMV tripartite genome, in which the CMV coat protein gene was replaced by the gene encoding α-1-antitrypsin [AAT] [9]. We deleted a region of the CMV RNA 1 leader sequence to ensure that the viral replicase was not able to replicate the truncated RNA 1 and since coat protein was lacking, infectious CMV was not generated thereby eliminating possible unwanted spread of the recombinant CMV. Furthermore, because one of the key CMV-encoded protein components of the viral replicase (1a) is under the control of a relatively tightly regulated chemically inducible promoter (the XVE inducible promoter [10]), recombinant viral amplicons were produced intracellularly only after addition of the inducer (β-estradiol). The high efficiency and specificity are among the major advantages of the XVE system, and thus it provides a potent tool for research in plant biotechnology.
Despite the advantage of having all CMV components on a single plasmid (e.g. ensuring that all CMV components are simultaneously introduced into the same cell) [9], the CMViva plasmid proved to not be easy for subsequent manipulation. Its size alone (28 kbp) made subsequent cloning manipulations difficult. Therefore, here we explored development of new CMV-based smaller-sized variants by separating genome components onto different plasmids to give a bipartite inducible (CMVin, CMV-based inducible system) and tripartite, autonomously replicating forms of CMV (CMVar, CMVbased advanced replicating system). We also assessed the effects of mRNA4 leader sequence variants and compared two CMV genotypes for their abilities to give in planta production of two proteins, the green fluorescent protein (GFP) and the Acidothermus cellulolyticus endo-1, 4-β-glucanase (E1), a cellulose degrading enzyme. This heat-stable, 56,000 MW well-studied endoglucanase has been produced previously in different species of transgenic plants [11][12][13], and is believed to have potential application for cellulose biomass conversion to sugars and use in biofuel production. Here we show that active E1 can be rapidly produced in non-transgenic plants by using agroinfiltration with the CMV-based systems. This reduces the time and energy required to generate transgenic plants and still gives the comparable yields of active E1 to those obtained previously by others.

Plants and photography
Three-week-old Nicotiana benthamiana plants and nine day old zucchini squash (Cucurbita pepo L. cv. Green Bush) plants were used for virus inoculations or agroinfiltration. Plants were photographed with a Cannon G6 digital camera equipped with a Tiffen Deep Yellow 15 filter. For photographing GFP expression, plants were illuminated with a hand-held long-wave UV lamp.

Cloning and plasmid construction
In order to develop CMVin (CMV-based inducible system) and CMVar (CMV-based advanced replicating system), the gene-of-interest was inserted into the coat protein coding region of CMV-Q RNA 3 (GenBank: M21464) [14] to give pCMVar RNA 3. We modified the CMV RNA 3 intergenic region, which also gives rise to the mRNA 4 leader sequence, by PCR primer tagging to introduce additional restriction enzyme sites for easier cloning. This was done by PCR amplifying the CP coding region using tagged forward primers (EATG for sequence 2, HATG for sequence 6, PHATG for sequence 8, and CPfwd for the wild type leader sequence, Table 1) and the reverse primer (CPrev listed in Table 1), and GoFlexi Taq DNA polymerase (Promega Corp., Madison, WI, U.S.A.). Amplified fragments were transferred to pGEM-T Easy (Promega Corp., Madison, WI, U.S.A.) and sequences were verified. Plasmids were then digested by Pst I and Tth111 I, and the desired fragment was transferred to pQA3 [9] (Additional file 1: Figure  S1). Then sequences containing the CaMV 35S (35S) promoter, RNA 3 and 35S terminator were PCR amplified using primers 35SPfwd and 35STrev (Table 1) and Pfu DNA polymerase (Stratagene, Agilent Technologies-Company, U.S.A), and the resulting fragments were ligated into the Sma I site of the mini binary vector, pCB301 [15] (Additional file 1: Figure S1). These were then used as the RNA 3 source for CMVar and CMVin variants. The higher producing constructs (containing the 2, 6, and 8 modified, and wildtype leader sequences of RNA4; Additional file 2: Figure S2) were selected for further experiments.
The E1 sequence (GenBank:HQ541433) used here was first codon-optimized for dicots and constructed to contain the rice alpha amylase (RAmy 3D, GeneBank: M59351) signal peptide at its N'-terminus, and a 6-His tag at its C' terminus (synthesized by DNA2.0, Menlo Park, CA, http://www.dna20.com, and provided as plasmid DNA pJL201:11772) ( [16], see Figure 1A). The green fluorescent protein (GFP) and E1 coding sequences were PCR amplified and cloned into the CP coding region of pCMVar RNA 3.
First, primer sets downstreamfwd and the Rna4wtrev, Rna42rev, Rna46rev, Rna48rev (listed in Table 1), were used for reverse PCR to remove the CP coding region (Additional file 3: Figure S3). The E1 gene was amplified by PCR using specific primers set (endoonlyfwd and endoonlyrev as listed in Table 1) and ligated into coat protein gene-deleted pCMVar RNA 3 by blunt end ligation (Additional file 3: Figure S3), yielding pCMVar E. The GFP coding sequence was PCR amplified from pCMViva GFP [9] using the specific primer sets (GFPfwd and GFPrev listed in Table 1), and cloned into the coat protein region of pCMVar RNA 3 using the same methods as for E1, resulting in pCMVar G (Additional file 3: Figure S3).
To generate the CMVar replicating constructs, CMV RNA 1 and RNA 2 segments (for CMV subgroup I and II) were PCR amplified using the specific forward and reverse primer sets (RNA1fwd and RNA12rev for subgroup II RNA 1, RNA2fwd and RNA12rev for subgroup II RNA 2, IRNA1fwd and IRNA12rev for subgroup I RNA 1, IRNA2fwd and IRNA12rev for subgroup II RNA 2, respectively, as listed in Table 1). The RNA 3 region of pCMVar RNA 3 was removed and replaced by the RNA 1 or 2 genome segments and gave I and II pCMVar RNA 1 and 2 (Additional file 4: Figure S4). The subgroup I RNA 1 and RNA 2 were originally from a California CMV [17], and the subgroup II RNA 1 and RNA 2 were from CMV-Q (GenBank:X02733 for RNA 1, X00985 for RNA 2, respectively). For the CMVin system, RNA 1 and 2 segments came from pDUXLR1R2 (pR1R2; [9] which includes the modified RNA 1 sequence. The monopartite inducible CMViva expression system, pCMV containing all three CMV genomic RNA segments in a single plasmid was used as control [9]. The plasmid, pCassQ123, containing all three CMV RNA segments in a single plasmid and each driven by the 35S promoter was a gift from Dr. ShouWei Ding, UC Riverside. In order to construct the CMV MP 33 amino acid deletion mutants, we used PCR and the specific primer set (33delfwd and 33delrev, listed in Table 1). PCR products were eluted from an agarose gel and self ligated to make pCMVar 33 G 2, 6, 8, wt and pCMVar 33E 2, 6, 8, wt variants, respectively (Additional file 5: Figure S5). Table 2 shows the names, activities and genotypes for the expression system variants used in this paper.

Agroinfiltration
Binary plasmids purified from E. coli cultures were transformed into Agrobacterium tumefaciens GV3101 or EHA105 cells using electroporation. Transformed A. tumefaciens cells were plated on Luria-Bertani plates containing Rifampicin (10 μg/ml) and Gentamycin (20 μg/ml) for GV3101 and Kanamycin (50 μg/ml) for specific constructs, and Rifampicin (10 μg/ml) and Tetracycline (10 μg/ml) for EHA105 and Gentamycin (20 μg/ml) for specific constructs, respectively. For agroinfiltration, a single colony was inoculated into 5 ml L-MESA media (100 ml LB broth, 2 ml 0.5 M MES (pH 5.7), 20 μl 0.1 M acetosyringone) and grown to an OD 600 of 1.0. Cells were harvested by centrifuging for 10 min at 3,500 g and resuspended in induction media (50 ml sterile dH 2 0, 0.5 ml 1 M MgCl 2, 1 ml 1 M MES (pH 5.7), 50 μl 0.1 M Acetosyringone), and allowed to sit at room temperature for 3 hrs before infiltration. When mixtures of A. tumefaciens cells were infiltrated into plants, cultures were prepared separately in induction medium and combined immediately before infiltration. For inoculating small sugar pumpkin plants, A. tumefaciens cells containing the constructs were infiltrated into N. benthamiana plants. Leaves were harvested 6 days after infiltration, and used for standard rub inoculation.

RNA extraction and realtime RT-PCR
Samples for RNA and protein extraction were harvested from infiltrated and non-infiltrated leaves at 6 days after infiltration. Total RNA was extracted using the RNeasy kit (QIAGEN Inc., U.S.A.) following the manufacturer's instructions. Complementary DNA (cDNA) synthesized from DNase-digested total RNA was used for reverse transcription using the RNA 3end primer as listed in Table 1 and SuperScript II Reverse Transcriptase, as described by the manufacturer (Invitrogen, Carlsbad, Modified leader sequences are located between the wildtype leader sequence and AUG start codon in subgenomic RNA4. Wt, wildtype leader sequence; 2+, modified #2 leader sequences addition to the wildtype leader sequence; 6+, modified #6 leader sequences addition to the wildtype leader sequence; 8+, modified #8 leader sequences addition to the wildtype leader sequence. CA, U.S.A.). Realtime PCR was performed using gene specific primers for each CMV RNA segment (realr-na1fwd and realrna1rev for RNA 1, realrna2fwd and realrna2rev for RNA 2, realrna3onlyfwd and realrna3onlyrev for RNA 3, realrna4onlyfwd and realrna4onlyrev for RNA 4, and real18Sfwd, real18Srev for endogenous 18S control, respectively as listed in Table 1). Real-time PCR was performed using SYBR Green PCR master mix (Applied Biosystems, Life Technologies Corporation, Carlsbad, CA, U.S.A.) in an ABI Prism 7500 Sequence Detection system (Applied Biosystems, Life Technologies Corporation, Carlsbad, CA, U.S.A.) under standard amplification conditions (95°C for 5 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min). The threshold cycle (C T ) is defined as the fractional cycle number at which the fluorescence exceeded the fixed threshold. Statistical analyses were performed using the Bonferroni (Dunn) t test using the SAS 9.1 program.

Protein extraction and immunoblotting
Proteins were extracted from leaves using protein extraction buffer (100 mM Tris-HCl (pH 8.0), 10 mM EDTA, 5 mM DTT, 150 mM NaCl, 0.1% Triton X-100, 1X protease inhibitor (Roche diagnostics, Germany) and tissue maceration using a bead-beater. Samples were centrifuged at 12,000 g for 20 min to remove cell debris, and protein concentrations were determined by Bradford assay using Coomassie Plus (Pierce, Thermo Scientific, IL, U.S.A.) with bovine serum albumin as the standard. Proteins were analyzed by SDS-PAGE in 12%

Results
Modified RNA4 leader sequences affect mRNA and coat protein levels To facilitate cloning desired genes into the CMV CP region, we first created a separate CMV RNA 3-based plasmid and modified the intergenic nucleotide sequence upstream of the CP ORF start codon to contain desired restriction endonuclease sites (Figure 1). This also resulted in changes to the mRNA 4 nucleotide leader sequence immediately preceding the AUG start codon and therefore could affect mRNA translation efficiency [18][19][20]. Therefore, we first compared the respective CMVin and CMVar variants for their ability to replicate and to express CMV CP within infiltrated leaves. We used realtime PCR to quantify levels of progeny RNAs 1, 2, 3, and 3 plus 4 (because RNA 4 is a subset of RNA 3 and is therefore difficult to differentiate from RNA 3).
When we compared the CMVinII variants with CMViva, all showed accumulation of genomic and subgenomic RNAs, and of CMV CP (Figure 2A). Because we are most interested in protein production from RNA 4, comparison of the RNA 3 & 4 data show that the modified variants (CMVinII 2, 6 and 8) all showed slightly more RNAs 3 and 4 than did the wildtype CMVinII, but none were higher than CMViva (Figure 2A). Although the levels of RNAs 3 and 4 were not statistically different, comparison of CMV CP accumulation showed more CMV CP was detected for CMVinII variants 2, 6, 8 and CMViva, than was for the wildtype CMVinII. CMViva and CMVinII variants both have 46 nucleotide deletions in the inducible CMV RNA 1 such that although it is transcribed and the resulting RNA 1 serves for translation to yield the 1a protein, the RNA 1 genome segment is not replicated as are genome segments RNAs 2 and 3 (Additional file 2: Figure S2). Because the 1a component of the replicase complex is under the control of a relatively tightly regulated chemically inducible promoter, the recombinant viral amplicons are only produced under induction conditions.
When we compared the CMVarII variants, all showed accumulation of all CMV RNAs, and so long as inocula contained all three CMV genomic RNAs, accumulation of RNAs 1, 2 and 3 was not significantly different for the different variants. However, realtime PCR analysis showed lower accumulation of RNAs 3 + 4 for CMVarII variant6 when compared to the monopartite pQ123 ( Figure 2B). All CMVarII variants showed much less CP accumulation than was seen for pQ123, and CMVarII 2, 6 and 8 showed less CP than did CMVarII wt ( Figure 2B). However, all CMVarII variants were able to initiate systemic infections in N. benthamiana and zucchini squash plants (Additional file 6: Figure S6). RT-PCR and nucleotide sequence analysis of the CMV progeny showed that the modified leader sequences were retained in RNAs extracted from systemically-infected leaves (data not shown).

CMVinII and CMVarII variants yield high GFP fluorescence
To assess foreign protein production we first cloned the gene for GFP into the CP coding region for the CMVinII and CMVarII variants. Non-transgenic N. benthamiana plants were infiltrated and leaves were examined for fluorescence at 6 and 10 days post-infiltration. At 6 days post-infiltration bright GFP fluorescence was seen in the infiltrated regions (Figure 3). In general, the regions of the leaves infiltrated with CMVinII and CMVarII variants showed very bright fluorescence. By 10 days postinfiltration, bright GFP fluorescence was observed for the CMVinII and CMVarII variants, regardless of leader sequence. Despite the fact that CMVarII variants producing the CMV CP spread within plants giving systemic infections, CMVarII variants producing GFP did not, and fluorescence was localized to the infiltrated areas. This is most likely because these variants do not produce CMV CP, which is known to be a determinant of CMV systemic spread in plants [21,22], and thus these infections were localized to the infiltrated regions of the treated leaves. We also tested another reporter protein, the red fluorescent protein (RFP), for expression using CMVinII and CMVarII and obtained essentially identical results to those shown for the variants expressing GFP (data not shown).

E1 was produced in plants using both CMVinII and CMVarII variants
Our intent is to develop easy-to-use CMV variants that give efficient production of desirable proteins in nontransgenic plants, including proteins with potential biofuel applications. Therefore, we next assessed E1 accumulation in leaves of infiltrated N. benthamiana plants. We first used immunoblot analysis to detect total E1 accumulation in leaves infiltrated with the CMVinII and CMVarII variants. E1 was detected in the infiltrated leaves for both CMVinII and CMVarII wild type variants ( Figure 4). The leader sequence CMVinII variants 6 and 8 showed higher E1 compared to the wildtype CMVinII. For CMVarII variants, the wild type showed higher E1 accumulation, but the 6 and 8 variants also gave good E1 accumulation (Figure 4). Interestingly, the intact E1 migrated as a ca. 72 KDa protein as shown in Figure 4 and Additional file 7: Figure S7, even though the calculated MW of E1 is 57.3 KDa, including the histidine tag and rice amylase signal peptide (See Figure 1A). Similar reports for anomalous E1 migration in SDS-PAGE have been previously reported [16,23].
The immunoblot experiments showed E1 protein accumulation but our interest is in production of enzymatically active E1. Therefore we used activity assays on plant extracts to estimate yields of active E1. In repeated experiments, the CMVinII E6 and E8 consistently yielded 8 to almost 15 fold higher relative E1 accumulation than wildtype CMVinII E, while the yield for CMVi-nII E2 was negligible. By contrast, the wildtype CMVarII E consistently yielded more E1 than the other CMVarII variants (Table 3). CMVarII E2 gave negligible E1 accumulation while CMVarII E6 and E8 showed low yields, but much less than wildtype CMVarII E. We had anticipated that CMVarII would give higher protein accumulation in plants because of RNA 1 replication, which is lacking in CMVinII (Additional file 2: Figure S2), but this proved not to be the case, the highest overall yields were obtained with CMVinII E 6 and 8. The CMVarII is easier to use since there is no requirement for adding the RNA 1 inducer, estradiol, and if we could achieve higher protein accumulation with CMVarII this would be our choice. Therefore we next attempted to increase CMVar-driven protein accumulation by two additional approaches: to increase CMV RNA replication and to increase CMV spread within plants.
Reassortant CMVarI variants yield more protein compared to CMVarII variants CMV is one of the world's most widespread plant viruses, and has many genetic variants which are primarily divided into the taxonomic subgroups I and II [5]. In general, subgroup I CMVs show more severe symptoms in plants than do subgroup II CMV isolates, which can show mild or even symptomless infections. This is suggested to be associated with the 2b protein (encoded by RNA 2) as a silencing suppressor [24], and effects can vary in different plant hosts [25]. Therefore, we generated and compared CMV subgroup I and II genomic reassortants for their abilities to give greater replication and protein production. All reassortants contained the same CMV subgroup II wildtype RNA 3 or variant constructs for GFP or E1. GFP fluorescence was brighter for all CMVarI variants compared with the respective CMVarII variants ( Figure 5). We next compared production of active E1 among the CMVarI and II E variants by immunoblotting (Additional file 7: Figure S7) and found that the reassortant wildtype CMVarI E gave more active E1 than did CMVarII E in side-by-side experiments (Table 3 and Additional file 7: Figure S7). By contrast, (See figure on previous page.) Figure 2 Relative RNA and coat protein accumulation levels for CMVinII and CMVarII variants having wt or modified RNA 4 leader sequences. After 6 days post-infiltration, infiltrated leaves were harvested and total RNAs and proteins were extracted as described in Methods. For CMVinII variants (see Table 2), estradiol was applied 18 hrs after infiltration using a cotton swab. For CMVinII variants, two leaves each from 3 plants for each treatment (n = 6) were used for 1 st , 2 nd biological trials, and three leaves each from 2 plants for 3 rd , 4 th trials (n = 6) yielding four different biological trials (total n = 24). For CMVarII variants (see Table 2), two leaves each from 3 plants for each treatment (n = 6) were used for analysis. These experiments were replicated three times (total n = 18). For immunoblot analysis, each lane was loaded with 15 μg of total protein. Real-time PCR reactions were performed using an ABI 7500 with gene specific primer sets. Relative Ct values were calculated to the respective 18S rRNA relative concentration. Relative folds were calculated to the repective wildtype CMVinII and CMVarII value, respectively. Group superscript letters next to the numbers represent different statistical groups, means with the same letter are not significantly different by the Bonferroni (Dunn) t test using the SAS 9.1 program. Panel (A) CMVinII variants; H , healthy uninfiltrated leaf; pR1R2, pCMVinII RNA 1 and RNA 2 only; CMVinII wt , wt leader; CMVinII 2 , #2 leader; CMVinII 6, #6 leader; CMVinII 8 , #8 leader; CMViva, pCMV (see Methods). Panel (B) CMVarII variants; H, healthy, uninfiltrated leaf; R1 + R2, pCMVarII RNA 1 and RNA 2 only; CMVarII wt , wild type leader; CMVarII 2 , #2 leader; CMVarII 6 , #6 leader; CMVarII 8, #8 leader; pQ123, pCassQ123 (see Methods).  Table 2) columns are indicated for both 6 and 10 days after infiltration. Rows show relative GFP fluorescence for leader sequence variants. A. tumefaciens cells containing respective plasmids were mixed just before leaf infiltration. CMVinII G plants then had estradiol induction treatments 18 hrs after infiltration. Photos were taken under long wavelength UV light. Figure 4 Immunoblot analysis of E1 produced in infiltrated N. benthamiana leaves. N. benthamiana leaves were co-infiltrated with A. tumefaciens containing plasmids corresponding to the CMVinII E and CMVarII E variants (see Table 2) shown. CMVinII E leaves were treated with estradiol 18 hrs after infiltration. Leaves were harvested 6 days post-infiltration. Soluble proteins were extracted as described in Methods and concentrations were determined by Bradford assay. Numbers above each lane correspond to the specific RNA 3 variant used. Each sample represents 15 μg total protein extract per lane. Prestained size marker (Benchmark, Invitrogen, Carlsbad, CA, U.S.A.) shows 80 kDa band. Numbers indicate E1 production as μg/g fresh weight. Average ± standard deviations are shown. Samples were collected 6 days post-infiltration, using 3 plants, 2 leaves from each plant (n = 6) for each biological trial. Results for three different biological trials are shown for each test. The identities of variants tested are given in Table 2. Group superscript letters next to the numbers represent different statistical groups, means with the same letter are not significantly different by the Bonferroni (Dunn) t test using the SAS 9.1 program.
CMVarI E variants 2, 6 and 8 gave very low E1 accumulation. However, wildtype CMVarI E gave relatively high E1 accumulation, similar to that for CMVinII E variants 6 and 8.

The MP C-terminal 33 amino acid deletion constructs showed increased yields compared to the intact MP constructs
CMV requires both the MP and CP for cell-to-cell movement in plants, both of which are encoded by RNA 3 [5,26]. Thus for both CMVar and CMVin variants, when foreign sequences are cloned into the CP coding region, there is no cell-to-cell movement due to lack of the CMV-encoded CP, and the desired recombinant proteins (GFP or E1) accumulate only in the initiallyinfected cells. However, it was shown previously that when the CMV MP was mutated so as to lack the Cterminal 33 amino acids, CMV infections were able to move cell-to-cell in plants even in the absence of CP [27]. Therefore, we deleted the MP C-terminal 33 amino acids and compared E1 and GFP accumulation in plants using the CMVarI and CMVarII variants. GFP fluorescence was high for all variants with the 33 amino acid truncated MP ( Figure 6). However, comparing E1 accumulation for all variants, the highest levels of active E1 were obtained for the 33 amino acid truncated CMVarI Ewt variant (Table 3). Our assays were for intact, enzymatically active E1, and we obtained yields up to 21 μg/g of active E1, corresponding to~0.4% of TSP. Furthermore, unlike for the wildtype MP variants, the CMVinII 33E variants 6 and 8 gave relatively low accumulation of E1.

Discussion
Several different viruses have been used for protein production in plants, and each has advantages as well as disadvantages [28][29][30][31][32]. Many of the "first generation" plant virus vectors [28,33] utilized whole plant systemic virus infections to give desired proteins. While many of these have proven to be very useful there are some significant drawbacks. Systemic infections can take several days to fully develop. Protein production is then asynchronous and yields can vary in different tissues [3]. Recombinant viruses also show size constraints for the inserted sequence, often coding sequences of only 1 kb or less (encoding a protein of only~35 kDa) can be Figure 5 Comparison of GFP fluorescence from CMVarI G and CMVarII G variants. N. benthamiana leaves were co-infiltrated with A. tumefaciens containing plasmids corresponding to the CMVarI and II G variants (see Table 2) as shown at Panel A. Numbers for leaves in panel at right indicate specific RNA 3 variant. Photographs were taken 6 days post-infiltration under UV light shown as Panel B.

Figure 6
Comparison of GFP fluorescence CMVarI 33 G and CMVarII 33 G MP deletion mutants. N. benthamiana leaves were co-infiltrated with A. tumefaciens containing plasmids corresponding to the CMVarI and II 33 G variants (see Table 2 inserted [34]. Then as the infection develops the viruses partially or completely excise the inserted recombinant sequence, leading to loss of the desired intact protein product [28,35]. Furthermore, some viruses (e.g. those with icosahedral capsids such as CMV) may have even more severe size constraints if RNA encapsidation is a requirement for development of the systemic infection. Then if the coding sequence for the desired protein is large, insertion into the viral RNA may preclude encapsidation, thereby preventing efficient spread. Recent progress in developing Agrobacterium tumefaciens delivered plant virus-based protein production systems has been made by several research groups using different plant viruses [31,34,36]. These plant virusbased amplicon systems offer many advantages including the fact that non-transgenic plants can be used, the desired protein production is rapid, the product can accumulate to high levels, and virus-based expression can be temporally regulated to be almost synchronous in all infiltrated areas. Because a majority of the infiltrated cells become simultaneously infected, virus movement to new cells is not necessary, encapsidation of recombinant RNAs is not an issue. These "second generation" virus-based systems also can retain larger foreign coding sequences and thus produce larger proteins in plants [34], here we produced enzymatically active 56,000 MW E1.
In our previous work, we used the estradiol-inducible, CMV-based CMViva to produce α anti-trypsin (AAT) in non-transgenic N. benthamiana plants [9]. CMViva has all three CMV genome components in one large 28 kbp plasmid, which, due to its large size is difficult to manipulate. Thus, here we took approaches to develop CMV-based inducible (CMVin) as well as autonomously replicating (CMVar) systems, both of which are more easily manipulated and might be able to give high accumulation of heterologous proteins in plants. First, we separated the CMV genomic RNA cDNAs onto two plasmids, one containing the RNA 1 and 2 replicationassociated genome components and the other containing the CMV RNA 3 genome segment. The CMV RNA3 component is rather small in size, 2.2 kb, and is easy to manipulate and to engineer to contain restriction enzyme sites to allow for easy removal of the CMV CP gene and replace it with any gene of interest. The desired restriction enzyme sites were introduced into the intergenic region of RNA 3. As expected, these altered the 50 untranslated leader sequence of the resulting mRNA (RNA 4). The sgRNA promoter (for RNA 4 transcription) is within the minus strand of RNA 3 and is recognized by the RNA-dependent RNA polymerase and mRNA transcription is initiated. For CMV-Q RNA3, the transcription initiation starts at nt position 1167 in the intergenic region, which is upstream of the modified leader sequences. Our analyses demonstrated that the RNA 3 modifications affected RNA 3 and RNA 4 accumulation, but showed even more unpredictable effects on resulting protein accumulation. It does not appear that these can be attributed only to start codon context [18,19] as the same construct (RNA 3) showed different protein yields whether the RNA was delivered using CMVin vs. CMVar.
In contrast to CMViva, both CMVinII and CMVarI and II variants require mixing A. tumefaciens cells containing different plasmids which are then co-infiltrated into plants and T-DNA from the different A. tumefaciens cells containing the CMV plasmids must be transferred to the same plant cell for the complete CMV amplicon. For CMVinII variants this is then followed by induction using estradiol, which resulted in high level accumulation of the proteins tested here (CP, GFP, E1). However, like for CMViva, the CMVinII RNA 1 deletion does not allow for its replication, only translation of the newly transcribed mRNA. Therefore we also developed the non-inducible autonomously replicating CMV-based system, CMVarI and II. Wildtype CMVar (expressing the CMV CP) replicated to very high levels and even caused systemic infections in plants. However, when genes for GFP or E1 were substituted for the CP gene, both proteins were produced in plants within the infiltrated areas, and quantitative analyses showed that high levels of proteins accumulated for both CMVin vs. CMVar, particularly at 6 days post-infiltration.
Although CMVinII E 6 and 8 variants gave slightly more active E1 in most experiments, CMVarI and II variants offers advantages in ease of use (e.g. no need to add the inducer) and thus, two additional approaches to improve accumulation of the desired protein product were investigated. Like most viruses having genomes composed of multiple segments, CMV genomic RNAs can be mixed (reassortment) to achieve genetic diversity [37,38], and this offers opportunities for using CMV to produce desirable proteins in different plant species, as has been demonstrated also by others [3]. Therefore, we generated CMV reassortant genotypes by substituting CMV subgroup I genomic RNAs 1 and 2 derived from a more virulent CMV, with the original CMV Q subgroup II RNA 3, giving CMVarI. Comparison of CMVarI and CMVarII G, E variants showed higher GFP and E1 for CMVarI G, E variants. However, the CMVarI Ewt showed higher E1 accumulation than did the corresponding CMVinII Ewt, but CMVinII E variants 6 and 8 gave the higher E1 accumulation than CMVar I 6 and 8 variants thus showing that reassortment alone was not sufficient.
As another alternative, we generated a MP C-terminal 33 amino acid deletion mutant. Cell-to-cell movement in CMV-infected plants requires interactions between the CP and MP [22]. Our CMV-based systems including CMViva, CMVarI and II and CMVinII are cell-to-cell movement deficient since they lack the CP and thus, desired recombinant proteins are produced only within infiltrated cells. However, previous workers demonstrated that the CMV MP C-terminal 33 amino acids are essential to recognize and interact with the CP [26]. When this region is deleted, the CMV infections can spread cell-to-cell even in the absence of CP [27]. In support of this the CMVarI and CMVarII 33 G variants showed high CMV-based GFP production ( Figure 6; and see [4]). When we created MP 33 amino acid deletion constructs and tested them, they showed increased production of not only GFP in CMVarI and II 33 G variants, but also of E1 in CMVarI and II 33E variants (Table 3, Figures 5 and 6), and the highest yields of active E1 were obtained using the CMVarI 33E variants.
Other workers have produced versions of E1 in various transgenic plants with gene expression driven by different promoters. For example, full-length E1 containing the catalytic domain, linker and carbohydrate binding domain has been previously produced in transgenic tobacco plants. Based on the resulting E1 activity, yields of up to 0.25% on average of total leaf soluble proteins were shown with Mac promoter, a chimeric promoter of the CaMV 35S and mannopine synthase gene [11]. Similar yields were shown also with CaMV 35S promoter [13]. In transgenic Z. mays seeds, the full-length E1 was produced using Glob-1 (Maize embryo-preferred globulin-1 promoter) and yields up to 6% TSP were obtained [39]. In transgenic rice (Oryza sativa) plants, 35S driven E1 lacking the carbohydrate binding domain but only containing the catalytic domain gave yields up to 4.9% TSP [40]. Thus, our yields of up to 0.4% TSP of intact E1 in nontransgenic N. benthamiana plants are similar to those achieved for intact E1 in transgenic tobacco, but less than those in more specialized systems. Furthermore, CMVin and CMVar-based production of the desired protein can be temporally regulated to give almost synchronous protein accumulation over a very short time period, even a few days.